Construction Classics

Knowing before building

One question, answered two ways, fourteen years apart. In 1947 you found out what an RF system would do by building half of one and measuring it. By 1961 you found out by writing an orbit code and running it. All four reports exist for the same reason: metal is expensive and a mistake in the magnet gap is expensive twice.

The first two are a matched pair, and finding the second half of it is why this page now has four documents. In September 1947 MacKenzie proposed the “three quarter wave” RF system and set out its theory with early model results. In December 1947, three months later and at the same laboratory, Anderson reported the half-scale model tests of that same system. The collection now holds the proposal and its verification. MacKenzie is candid in the introduction about which half of his own report he expects to last:

By the time it is available, the experimental results quoted in the report will be superseded, it is hoped, by much more recent and encouraging data. However, the theory should apply, and may help to clarify some of the improvements found by empirical methods.

The design he is arguing for puts the rotating condenser about a third of the way along the dee line from the shorted end, which gets it outside the magnetic field — no eddy current losses, room to build it ruggedly, and, as he notes, the ability to service it without much danger of an overdose of radiation from a ‘hot’ dee. He states the costs just as plainly: bulkier than a condenser mounted on the dee, more power, and many more possible modes of oscillation. A 2.5:1 frequency swing was achieved against a theoretical 3:1, and the limit was how to couple the oscillator tube across the range.

Why these documents can be hosted

AECD-1850 is a declassified AEC report — manuscript September 1947, declassified March 1948, issued by the Commission's Technical Information Branch at Oak Ridge — the same basis as the MDDC memoranda hosted elsewhere here. Its 24 pages carry no copyright notice, no residual classification marking and no journal-reprint marking.

UCRL-31 is a University of California Radiation Laboratory report under AEC contract W-7405-Eng-48, carrying no copyright notice, no classification marking and no journal-reprint marking — the same basis as every other UCRL report in this collection.

The two MSU reports need a different argument, and it is worth stating rather than borrowing. MSUCP is a university series whose footnote reads “Research supported in part by U. S. Atomic Energy Commission Contract AT(11-1)-872.” Supported in part, by university employees, is not the same as a work produced under contract by a laboratory operator, so the US-Government-work argument is weaker here than it is for an Oak Ridge or Berkeley report. The load-bearing basis for these two is therefore pre-1978 distribution with no copyright notice — verified absent from both scans — with the AEC sponsorship and the standard government-sponsorship disclaimer as support rather than as the claim. Determinations made August 2026, by reading the scans.

1947: build half of it

UCRL-31 tests a half-scale model of the RF system for the 184-inch cyclotron — the three-quarter-wave system for frequency-modulated machines. It follows a preliminary report by Kenneth MacKenzie from September 1947 and states its own scope precisely: to present the apparatus and the results of tests performed on the half-scale model since that date.

The sentence to take away is the one about what happens next: Further tests will be made on the full size r.f. system prior to its actual installation in the 184-inch cyclotron. That is a three-stage validation chain — half-scale model, then full-size system on the bench, then installation — and each stage exists to catch what the previous one could not. Berkeley did not model instead of testing, or test instead of modelling.

MacKenzie’s name is the connective tissue here. He co-authored the already-hosted UCRL-64, the design history of that same 184-inch RF system, and MDDC-1045 on the 37-inch FM conversion that preceded it. Read in order — MDDC-1045, UCRL-31, UCRL-64 — the three give one RF system from prior art through scale model to as-built design.

Scale modelling is the method the collection keeps returning to. NYO-780 ran magnet models at 2, 6 and 9 inches before committing forgings; TID-454’s sixth report is a three-quarter-scale model oscillator; ORNL-2648 established the model-magnet method for AVF machines. UCRL-31 is the RF instance, and the earliest.

1961: compute it instead

The two MSU reports are the same instinct with a computer behind it, from the Michigan State cyclotron project as it was being designed.

MSUCP-12 is the one to read first, and the more useful of the two at small scale: analytic solutions for the electric field and potential of an idealized dee geometry, obtained by transforming the dee cross-section into a geometry whose field is already known, and tabulated for several dee arrangements. Nineteen pages. Because the result is analytic rather than a machine-specific field map, the formulas do not care how big your dees are — which is exactly why it was worth acquiring for a site about small machines. It builds on earlier work on fields within cyclotron dees, with annotations and corrections stated as such.

MSUCP-9 is Blosser and Gordon on resonant extraction optics for a medium-energy three-sector cyclotron, studied by orbit tracking — the Fixed Point Code and the General Orbit Code, run on MISTIC, with a linear transfer-matrix method for the well-behaved cases. It is candid about the limits of the approach: reading design conclusions out of orbit plots and phase diagrams “is not however an easy” matter, so both static and accelerated orbit behaviour had to be computed. Blosser is the same author as ORNL-2648, three years earlier at Oak Ridge.

Why the pairing matters now

An amateur builder in 2026 has both methods available and cheaper than either was then. The computation is free — see modeling tools for the open-source field solvers and orbit codes that do what MISTIC did. The physical model is a weekend of aluminium and a network analyser.

What these reports supply is the judgement about which to use. UCRL-31 models the RF system, where the awkward quantities are contact resistance, joint behaviour and standing-wave patterns in real conductors — things a field solver still models badly. The MSU reports compute orbits and electrostatic fields, where geometry is exact and measurement is hard. The 1947 answer and the 1961 answer are not competing; they were applied to different subsystems, and mostly still should be.

Design rules extracted from these documents

47 Design Guide rules come from these reports. Representative examples:

• dg-999 — retire RF-system risk with a scaled electrical model before cutting full-size metal
• dg-1003 — tune with every electrode in place — the dummy dee alone detuned the model by ~9%
• dg-1002 — three parasitics set a dee system's resonant range and deserve first attention
• dg-1011 — a scale model's known infidelities must be listed alongside its results
• dg-1028 — the peak accelerating field saturates at V₀/h — set by the aperture, not the gap
• dg-1030 — do not use the parallel-plate V/d estimate for dee-gap fields
• dg-1031 — Table 1 is a ready-made verification dataset for anyone implementing the solution

Used by the modeling tools and dee RF coupling pages.